Environment:Scikit learn Scikit learn Python Runtime Environment

Knowledge Sources	scikit-learn pyproject.toml _min_dependencies.py
Domains	Infrastructure, Machine_Learning
Last Updated	2026-02-08 15:00 GMT

Overview

Python 3.11+ environment with NumPy, SciPy, joblib, and threadpoolctl as core runtime dependencies for scikit-learn.

Description

This environment defines the core runtime requirements for using scikit-learn. It is a CPU-based Python scientific computing stack. Scikit-learn requires Python >= 3.11 (supporting 3.11, 3.12, 3.13, and 3.14) and runs on Windows, POSIX, Unix, and macOS operating systems. The library is implemented as a mix of Python and Cython (compiled C extensions), requiring a proper build step for source installations. Pre-built wheels are available for common platforms.

Usage

Use this environment for any scikit-learn workflow, including supervised classification, data preprocessing, hyperparameter tuning, cross-validation, and ensemble model building. This is the mandatory baseline prerequisite for all scikit-learn implementations.

System Requirements

Category	Requirement	Notes
OS	Windows, POSIX, Unix, macOS	All major platforms supported via CPython
Python	>= 3.11	CPython only; 3.11, 3.12, 3.13, 3.14 are classified
Architecture	x86_64, aarch64, wasm32/64	32-bit and PowerPC marked non-deterministic for numerical tests
Build System	Meson + Cython	Required only for source builds, not pip install from wheel

Dependencies

System Packages

C compiler (gcc/clang/MSVC) — only for source builds
OpenMP runtime library — for parallel computation in Cython extensions

Python Packages (Core Runtime)

`numpy` >= 1.24.1
`scipy` >= 1.10.0
`joblib` >= 1.3.0
`threadpoolctl` >= 3.2.0

Python Packages (Build)

`cython` >= 3.1.2
`meson-python` >= 0.17.1
`numpy` >= 2 (build-time requirement, differs from runtime)
`scipy` >= 1.10.0

Python Packages (Optional)

`matplotlib` >= 3.6.1 — for plotting and visualization displays
`pandas` >= 1.5.0 — for DataFrame support in data loading and output
`polars` >= 0.20.30 — alternative DataFrame support
`pyarrow` >= 12.0.0 — for Arrow-based data handling
`scikit-image` >= 0.22.0 — for image processing examples
`seaborn` >= 0.13.0 — for enhanced plotting
`plotly` >= 5.18.0 — for interactive visualizations

Credentials

No credentials are required for core scikit-learn functionality. Network-dependent dataset loaders (e.g., `fetch_openml`) require internet access but no API tokens.

The following environment variables are available for configuration (not secrets):

`SKLEARN_ASSUME_FINITE`: Skip NaN/Inf validation (default: False)
`SKLEARN_WORKING_MEMORY`: Memory limit in MiB for chunked operations (default: 1024)
`SKLEARN_SEED`: Set RNG seed for test reproducibility

Quick Install

# Core installation (from PyPI wheel)
pip install scikit-learn

# With common optional dependencies
pip install scikit-learn matplotlib pandas seaborn

# For development (source build)
pip install numpy scipy cython meson-python
pip install --no-build-isolation -e .

Code Evidence

Python version requirement from `pyproject.toml:15`:

requires-python = ">=3.11"

Core dependencies from `pyproject.toml:9-14`:

dependencies = [
  "numpy>=1.24.1",
  "scipy>=1.10.0",
  "joblib>=1.3.0",
  "threadpoolctl>=3.2.0",
]

Minimum dependency constants from `sklearn/_min_dependencies.py:10-15`:

NUMPY_MIN_VERSION = "1.24.1"
SCIPY_MIN_VERSION = "1.10.0"
JOBLIB_MIN_VERSION = "1.3.0"
THREADPOOLCTL_MIN_VERSION = "3.2.0"
PYTEST_MIN_VERSION = "7.1.2"
CYTHON_MIN_VERSION = "3.1.2"

Build verification from `sklearn/__check_build/__init__.py:51-54`:

try:
    from sklearn.__check_build._check_build import check_build  # noqa: F401
except ImportError as e:
    raise_build_error(e)

Platform detection from `sklearn/utils/fixes.py:26-27`:

_IS_32BIT = 8 * struct.calcsize("P") == 32
_IS_WASM = platform.machine() in ["wasm32", "wasm64"]

Common Errors

Error Message	Cause	Solution
`ImportError: It seems that scikit-learn has not been built correctly`	C extensions not compiled	Run `pip install --no-build-isolation -e .` for source installs
`ImportError: numpy>=1.24.1 is required`	NumPy version too old	`pip install --upgrade numpy>=1.24.1`
`ModuleNotFoundError: No module named 'sklearn'`	scikit-learn not installed	`pip install scikit-learn`
`{caller_name} requires matplotlib`	Optional dependency missing	`pip install matplotlib`
`{caller_name} requires pandas`	Optional dependency missing	`pip install pandas`

Compatibility Notes

32-bit systems: Some numerical tests are marked non-deterministic on 32-bit platforms due to floating-point precision differences.
PowerPC (ppc/powerpc): Tests marked non-deterministic due to numerical instability on these architectures.
WebAssembly (wasm32/wasm64): Detected and handled via `_IS_WASM` flag; some features may have limited support.
Windows: NumPy array representation differs; doctests may be skipped on Windows platforms.
macOS: Requires `KMP_DUPLICATE_LIB_OK=True` environment variable (set automatically) to handle multiple OpenMP libraries.

Related Pages

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment